Market liquidity risk is the potential loss, because assets cannot be sold at the price previously thought. Although evidence suggests that liquidity effects are significant, they often remain neglected in practical risk management. One of the reasons is the limited scientific research in the area of liquidity risk measurement.
This thesis provides an up-to-date overview on market liquidity risk research. It covers all aspects of market liquidity that are relevant to risk management as well as existing liquidity risk models.
The empirical analysis is based on weighted spread, a relatively new liquidity measure, which can be extracted from the limit order book of electronic exchanges. A unique, representative sample of weighted spread allows to provide estimates on the effect of order size on liquidity costs, as well as the dynamics, distributional characteristics and cross-sectional structure of this liquidity measure.
The thesis also proposes two new liquidity risk models. The modified liquidity risk model introduces a new way to account for the non-normality in liquidity with the help of the Cornish und Fisher (1937)-approximation. The empirical net-return model based on weighted spread analyzes the use of the weighted spread liquidity measure in risk measurement.
Both models are tested empirically in daily data. The modified liquidity risk model implemented with the bid-ask-spread proves to be superior to the standard model of Bangia et al. (1999). Common backtests by Kupiec (1995) demonstrate that risk is forecasted with much higher precision when non-normality is taken into account via the proposed Cornish-Fisher approximation.
With the help of the empirical net-return model, I find that liquidity risk strongly increases with the size of the position. The impact of liquidity on risk is significant - even at 10-day horizons. Liquidity risk models neglecting this effect must necessarily underestimate total risk. Further, the correlation between liquidity and return is significant and reduces the liquidity impact by about 50 % compared with the standard assumption of perfect correlation. These results are robust to change in risk measure, effects of time variation as well as portfolio diversification.
A final test runs a performance benchmark of nine different liquidity risk models implementable in daily data, including the new propositions. I find that available data is the main driver of model preciseness. Models with extensive data from the limit order book generally outperform. My new propositions, modified add-on with weighted spread and empirical net-return with weighted spread as well as Giot und Grammig (2005), are all recommendable. The first model delivers precise results most consistently. If only transaction data are available, the model by Cosandey (2001) performs best. With bid-ask-spread data the proposed modified add-on model with bid-ask-spread achieves superior results.
Overall, this thesis underlines the usefulness of the weighted spread measure in liquidity risk modeling. If the analyzed structure of liquidity costs, i.e. non-normality as well as increase with order size, is properly integrated, the preciseness of risk forecasts can be greatly improved. The new model contributions prove to be particularly helpful in practical risk management.
«
Market liquidity risk is the potential loss, because assets cannot be sold at the price previously thought. Although evidence suggests that liquidity effects are significant, they often remain neglected in practical risk management. One of the reasons is the limited scientific research in the area of liquidity risk measurement.
This thesis provides an up-to-date overview on market liquidity risk research. It covers all aspects of market liquidity that are relevant to risk management as well a...
»
Login
BibTeX